Anthropic's Mythos AI model, which the company restricted from public release citing hacking risks, was accessed without authorization by a small group that exploited credentials from a contractor working with Anthropic. The core facts are confirmed across Bloomberg, Reuters, BBC, and TechCrunch, but Whittaker's framing oversimplifies how access occurred—the group didn't simply 'find it' on Discord but exploited existing contractor access. Anthropic is investigating the breach and the incident raises real questions about access control for powerful security tools, though the 'fucking joke' framing substitutes emotion for analysis of what actually went wrong in their security model.